⚡ WebGPU Benchmark Results (101.20x speedup)
#59
by
Xenova
- opened
| Batch Size | WASM (fp16) | WebGPU (fp16) |
| 1 | 1073.10 | 12.80 |
| 2 | 2146.10 | 178.50 |
| 4 | 4352.90 | 265.60 |
| 8 | 8749.50 | 337.20 |
| 16 | 17610.90 | 286.80 |
| 32 | 35285.90 | 631.10 |
| 64 | 70255.60 | 694.20 |
- Model: Xenova/all-MiniLM-L6-v2
- Tests run: WASM (fp16), WebGPU (fp16)
- Sequence length: 512
- Browser: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36
- GPU: vendor=nvidia, architecture=turing, device=, description=
Wow!
